Inspiration
Families record grandparents telling stories, then the recordings sit unplayed in a folder, apart from the photos they describe. A StoryTerrace survey found nearly three in four people had a relative die before their stories were written down.
What it does
Voice Story Albums is a UGOS Skill for UGREEN AI NAS. From one recording it:
- transcribes the recording on the NAS;
- lists each story moment (who, where, when, scene), each cited to a transcript sentence;
- searches the photo library using only what was actually said, shows a preview, and creates one album per confirmed moment;
- drafts a family story (optionally translated), with a Told but not pictured section for moments that have no photo;
- saves a story map and tags every file with one shared tag.
It never fills a gap with "similar" photos or guesses a year that was not spoken.
How we built it
The Skill is orchestration only: SKILL.md delegates in order to knowledge-agent, document-agent, photo-agent and file-agent (ug-knowledge, ug-document, ug-photo, ug-filemgr). Each step uses only the real result returned by the step before it, and every write waits for platform approval. It was built with the official UGOS Skill Creation Tool (uliya-skill-kit 0.7.7) and passes its static checker with 0 errors and 0 warnings.
Challenges
There is no "add to existing album" action, so the design makes one album per moment. Photo search has to stay strictly inside spoken filters, which means the Skill must ask before turning a decade into a date range.
What's next
If selected as a finalist: test an English recording, a Mandarin recording with an English story copy, and a moment with no matching photos on the assigned iDX6011 Pro, and report measured transcript accuracy and match rates.
Disclosure
AI tools used: Claude (Anthropic) for research, drafting the Skill files with the official creation tool, and drafting text that I reviewed. No third-party components.
Built With
- markdown
- ugos
- ugreen-ai-nas
- uliya
- uliya-skill-kit
